Multiscale optical Ca2+ imaging of tonal organization in mouse auditory cortex

Abstract

Spatial patterns of functional organization, resolved by microelectrode mapping, comprise a core principle of sensory cortices. In auditory cortex, however, recent two-photon Ca2+ imaging challenges this precept, as the traditional tonotopic arrangement appears weakly organized at the level of individual neurons. To resolve this fundamental ambiguity about the organization of auditory cortex, we developed multiscale optical Ca2+ imaging of unanesthetized GCaMP transgenic mice. Single-neuron activity monitored by two-photon imaging was precisely registered to large-scale cortical maps provided by transcranial widefield imaging. Neurons in the primary field responded well to tones; neighboring neurons were appreciably cotuned, and preferred frequencies adhered tightly to a tonotopic axis. By contrast, nearby secondary-field neurons exhibited heterogeneous tuning. The multiscale imaging approach also readily localized vocalization regions and neurons. Altogether, these findings cohere electrode and two-photon perspectives, resolve new features of auditory cortex, and offer a promising approach generalizable to any cortical area.

Figure 1. Transcranial responses to SAM tones in GCaMP3 mice

(A) Transcranial Ca²⁺ imaging layout. Speaker emits SAM (sinusoidal amplitude modulated) tones to right ear of head-fixed, unanesthetized mouse. 470-nm excitation illuminates thinned skull over left auditory cortex (gray circle). 525-nm emission collected by CCD camera. (B) Average transcranial fluorescence image from exemplar mouse, expressing GCaMP3 under Emx1-Cre driver line. Vasculature landmarks are low signal. (C) Ca²⁺ activity induced by SAM tones at three regions marked in panel B. Upper 3 rows, single-trial fluorescence responses. SAM tones delivered every 2 secs, covering 5-octave range in random order. For convenience, traces here sorted by increasing frequency, as labeled on bottom. Original baseline-corrected signal in gray and sparse-encoding waveform in black. (D) Ca²⁺ activity over entire auditory region during presentation of low, middle, and high-frequency tones. Images displayed in grayscale format (equivalent ΔF/F_O scale bar, lower right) after processing by sparse-encoding algorithm and deblurring. Loci of strongest response to low (L) and high (H) frequency tones marked on 3 and 30 kHz images, respectively. Dorsal-caudal spatial scale bar on lower left denotes 300 μm along each dimension. (B–D) Data from same mouse.

Figure 2. Formation of transcranial map of auditory cortices

(A) Single-trial transcranial fluorescence responses during SAM tones of various frequencies (x axis) and sound attenuations (y axis, left). SPL intensities, y axis on right, coarsely corrected for speaker calibration. Images as in Figure 1D, but plotted as inverted grayscale format (equivalent ΔF/F_O scale bar, lower right). Jagged thin black line indicates approximate threshold, lowest sound level eliciting responses at given frequency. Signals below threshold reflect spontaneous activity. Mouse expressing GCaMP3 under Syn1-Cre driver line. (B) Best-frequency map for the same experiment. Each pixel plots a color-map readout of weighted average of 3 frequencies eliciting largest responses at threshold sound level. Color map (lower right) associates best frequencies (y axis) with specific colors (e.g., blue = low, red = high); color saturation (x axis) denotes strength of best-frequency responses. Low- and high-frequency hotspots (L and H labels, respectively), determined as local center-of-mass coordinates calculated over low- and high-frequency islands, serve as landmarks throughout. Dorsal-caudal spatial scale bar on lower left denotes 300 μm along each dimension. (C) Transcranial response maps as in panel B, for 2 other mice (M2 and M7). Slight differences exist in colors and precise layout of landmarks, but overall features and low-to-high frequency gradients are preserved. Mice expressing GCaMP3 under Emx1-Cre (M2) and Syn1-Cre (M7). (D) Average of 7 transcranial response maps, including those in panels B–C, demonstrating conserved canonical layout across mice. Landmark-based registration morphs each individual map onto a common coordinate system before averaging (see Figure S5). (E) Global map from panel B, with dark-thick-line paths of maximally responsive loci (local center-of-mass) with increasing frequencies. Trajectories derived from solid arrows in panel F. (F) Coarse schematic of mouse auditory cortex, from transcranial imaging of mouse M1. Solid lines with arrows, best fits to data; dashed lines with arrows, poorer fits. Details in Figure S4.

Figure 3. Tonal tuning of exemplar neuron residing within low-frequency pole of AI

(A) Registration of neuron to transcranial map. Left subpanel, high- and low-frequency landmarks (H and L) obtained via transcranial imaging (scale bar, 500 μm). Box registers overall field of view for neuron-by-neuron imaging under two-photon microscopy. Middle subpanel, portion of actual field of view acquired under scanning two-photon microscopy (scale bar, 30 μm), reporting time-averaged fluorescence. Box registers individual neuron to be scrutinized in right subpanel. Right subpanel, change in fluorescence of this neuron between quiet (top) and active (bottom) periods (scale bar, 10 μm). GCaMP3 under Syn1-Cre driver line. (B) Single-trial fluorescence as a function of time from neuron in panel A (right subpanel). Time-registered blips show frequency and timing of 30 randomly delivered SAM tones logarithmically spaced between 3 and 48 kHz. (C) Fluorescence responses from same neuron, after sorting for tone presentation in order of increasing frequency (left to right). Each row relates to stimuli presented at different levels of sound attenuation. Results of individual trials plotted in gray and averaged responses across multiple trials in black. Strong and sharp tuning to low frequencies of 4–5 kHz. (D) Frequency-response area (FRA) for this neuron, confirming low-frequency tuning. Intensity displays estimated spike rate, gauged by deconvolving and thresholding traces (Figure S7).

Figure 4. Spatially co-localized AI neurons show sharp tuning to similar tone frequencies

(A–C) Neuron-by-neuron responses in low-frequency AI, encompassing same field as Figure 3. GCaMP3 under Syn1-Cre. (A) Landmarks from transcranial imaging, with registered two-photon imaging field. Same field and format as in Figure 3A, but two-photon field here encompasses modestly larger region (scale bar, 30 μm). Colored circles, best frequency of sound-responsive neurons, with color map at bottom (blue = low frequency, red = high frequency). Thicker circles delineate neurons whose responses appear in panel B. (B) Activity of 5 neurons marked in panel A; format as in Figure 3C. Neuron 3 is same as shown in Figure 3. Sound attenuation of −20 dB was used throughout panels B and H. FRAs on right follow format in Figure 3D. FRAs scaled to 24.6 events/sec. (C) Top subpanel, FRA averaged from all active neurons in entire field, showing considerable population tuning to low frequency stimuli. Scaled to 5.64 events/s. Middle subpanel, cumulative distribution of best-frequency spread (ΔBF, in octaves here and throughout) between all pairs of tuned neurons in field. Bottom subpanel, cumulative distribution of sharpness of tuning metric (Q factor) for all tone-responsive neurons in field, substantiating narrow frequency preference within individual neurons. Vertical dashed lines and symbols delineate mean values. (D–F) Single-neuron responses in mid-frequency field of AI, where tone-responsive neurons demonstrate clear tuning to similar middle frequencies. Format as in A–C, but from different mouse, with −40 dB sound attenuation. GCaMP3 under Syn1-Cre. Neurons 1–4 in panel E exhibit mid-frequency tuning. Fields often contained neurons with activity not driven by tones, as illustrated by neuron 5 in panel E (white-circled neuron in panel D). Individual FRAs in panel E scaled to 7.25 events/sec, and population FRA in panel F scaled to 1.21 events/sec. (G–I) Neuronal responses in high-frequency field of AI (or UF), where tone-responsive neurons demonstrate tuning to similar high frequencies. Format as in A–C; different mouse than in above panels. Data shown here (but not elsewhere) are from bulk-loaded Fluo-2 chemical-fluorescent dye, illustrating overall similarity to results obtained from GCaMP3. Individual FRAs in panel H scaled to 33.11 events/sec, and population FRA in panel I scaled to 4.90 events/sec.

Figure 5. Broad frequency tuning and diverse best frequencies in co-localized AII neurons

Exemplar low, middle, and high frequency AII fields, format as in Figures 4A–4C. (A–C) Neuron-by-neuron responses in low-frequency AII field, from mouse expressing GCaMP3 under Syn1-Cre. Neurons exhibit overall preference for low frequencies (panel B), but with broad frequency responsiveness and diversity of best frequencies. Broad population FRA in panel C supports this trend. Individual and group FRAs scaled to 25 (B) and 2.62 events/sec (C). Panels B, E, and H used sound attenuation of −20 dB. (D–F) Neuron responses in mid-frequency AII field, from another mouse expressing GCaMP3 under Syn1-Cre. Broad frequency tuning to divergent best frequencies (panel E), confirmed by population metrics (panel F). Note x-axis break for ΔBF to allow full display of larger ΔBFs. Individual and group FRAs scaled to 15.5 (E) and 2.66 events/sec (F). (G–I) Responses in high-frequency AII field, from mouse with GCaMP3 under Emx1-Cre. Moderately sharp tuning, albeit with greater ΔBF than in corresponding high-frequency areas of AI (cf., Figure 4G–I). Individual and group FRAs scaled to 17.5 (H) and 6.09 events/sec (I).

Figure 6. Contrasts in tonotopic organization of mouse AI versus AII cortex

(A–C) Well-resolved tonotopic gradient via two-photon imaging of individual neurons within AI. (A) Map of all two-photon imaging fields (squares in translucent ovoid) characterized along ventrodorsal tonotopic axis of AI, as resolved in transcranial maps. Each field registered onto a single canonical map (Figure S6 and Online Experimental Procedures). (B) Best frequencies of neurons versus normalized distance along AI axis, where the ‘0’ coordinate corresponds to the low-frequency pole in AI, and the ‘1’ position marks the high-frequency landmark atop AI (i.e., UF). Each black symbol marks outcome of one neuron, and ensemble of points encompasses all tone-responsive neurons from all fields in panel A (squares). Squares in panel B plot field averages. Good linear correlation (r = 0.88, 154 neurons, p< 0.001 via Pearson’s analysis) supports strong tonotopy in AI. Normalized distance on AI axis determined by orthogonal projection onto linear axis between low- and high-frequency AI poles, determined in transcranial maps for given neurons. Fields within 300 μm of the axis included. Red symbols, data from single larger field of view, obtained with 25× objective (Figure S8B–D). (C) Summary of co-tuning and tuning sharpness for all fields in panel A. Top, cumulative distribution of BF spread (ΔBF), as in Figure 4C. Bottom, cumulative distribution of sharpness of tuning (Q) averaged across all neurons within individual fields. Black-dashed vertical lines and symbols indicate mean values. (D–F) Weaker tonotopic gradient in AII, same format as in panels A–C. (D) Squares register fields residing near tonotopic axis of AII (translucent ovoid). (E) Best frequency plotted versus normalized position along AII tonotopic axis demonstrates poorer but significant correlation (r = 0.54, 143 neurons, p< 0.001 via Pearson’s analysis). (F) Weaker co-tuning and tuning sharpness in AII versus AI. Top, cumulative distribution of BF spread for all AII fields in panel D (black with yellow shading). Significant right shift of distribution (p< 0.042, 2-tailed Mann-Whitney, performed on metrics from 13 AI versus 16 AII fields) compared with AI distribution (fit reproduced in gray, for reference) indicates greater diversity of best frequencies at specific AII locales. Bottom, cumulative distribution of Q for all AII fields in panel D (black with yellow shading). Significant left shift of distribution (p< 0.02, 2-tailed Mann-Whitney, performed on metrics from 13 AI versus 16 AII fields) compared with AI distribution (fit reproduced in gray, for reference), showing decreased sharpness of tuning in AII. Vertical dashed lines and symbols, mean ΔBF and Q values for AII (yellow) and AI (cyan). (B–C, E–F). Data from mice expressing GCaMP3 under Syn1-Cre (n = 9) and Emx1-Cre (n = 8), and from mice bulk loaded with Fluo2 (n = 10).

Figure 7. Tone-insensitive sector along AI/AII border

(A) Comparison of low- and high-frequency poles of global transcranial map (from Figure 2B) to electrode-based map (Guo et al., 2012). Landmarks (L and H) translated without scaling or rotation to register low- and high-frequency loci. Ventral landmarks correspond to regions where it is difficult to assign best frequencies using electrodes (dots, indeterminate responsiveness). (B) Features of primary fields by multi-scale imaging. Global map annotated from Figure 2F. (C) Tone-insensitive region (dashed sector) identified by global transcranial maps from representative experiment. Left subpanel plots fluorescence response averaged across multiple SAM tone frequencies spanning entire mouse receptive range, while right subpanel plots same in response to a set of ten vocalizations (Figure 8A). Dashed oval denotes region insenstive to SAM tones. Both sets of stimuli presented at 20 dB attentuation. GCaMP3 expressed under Syn1-Cre. (D) Ratio of vocalization to tone responses, from same mouse as in panel C. Green regions correspond to pixels favoring vocalizations (by 4:1), and blue regions to those favoring SAM tones (by 1:4). Dashed oval same as in panel C.

Figure 8. Vocalization-selective neurons identified by multiscale Ca²⁺ imaging

(A) Spectrograms of ten adult vocalizations, adapted from online database (Grimsley et al., 2011). Silent periods removed for clarity. (B) Global map for this mouse identifies tone-insensitive region highlighted by magenta oval (left subpanel). High-frequency AAF region (rectangle) overlaid on color map showing ratio of vocalization to tone responses (middle subpanel). Right subpanel displays corresponding two-photon field (scale bar, 30 μm). GCaMP3 under Syn1-Cre, another mouse than in Figure 7. (C) Single-neuron responses to vocalizations. Single trials (top three rows) illustrate responses for three neurons identified in panel B. Bottom row plots average of multiple trials (dark trace, with individual trials in gray). (D) Single-neuron response to SAM tones, format as in panel C except only one trial is shown. No neurons in this field responded below 24 kHz, so traces are truncated.